quantitative risk and trading examplesuserhome.brooklyn.cuny.edu/ohadjiliadis/talksteven.pdf ·...

Quantitative Finance CareersRisk Examples

Trading ExamplesInterview Advice and Useful Skills

Quantitative Risk and Trading Examplesand Finding Your First Job

Steve Taylor

Morgan Stanley (Fixed Income Division)

Brooklyn College: Sept. 16th 2014

Steve Taylor Quantitative Risk and Trading Examples



Table of contents

1 Quantitative Finance CareersBanksHedge Funds/Prop ShopsFund Advisors/Asset Managers

2 Risk ExamplesRisk MetricsThe GreeksRegulatory Requirements

3 Trading ExamplesFinding an Optimal StrategyUsing NLP to Construct Market SignalsMaking Markets

4 Interview Advice and Useful SkillsSkills to DevelopInterview Advice




BanksHedge Funds/Prop ShopsFund Advisors/Asset Managers

Quants in Banks

Profile: Large institutions: 10K - 200K employees, LargeAUM: 10B – 2T

Investment Banks: Barclays, Goldman, Morgan Stanley, UBS

Commercial Banks: BAML, Citi, HSBC, JPM Chase

Many Types of Quants in Banks:

Pricing/Modeling, Research, Risk, Strats,✘✘✘✘Trading





Bank Quant Daily Workflow

Pricing/Modeling: Develop models for modeling dynamicquantities, including interest/FX rates, stock/bond prices, andcredit profiles of products (positions are in decline after crisis).

Risk: Develop models to understand risk profiles of financialproducts that the bank holds. Develop portfolio risk metricsincluding VaR, CVaR, etc. Implement Dodd-Frank regulation.

Strats: Jack of all trades quants responsible for the quantwork of a specific desk. Work with traders, IT, risk, andbusiness managers to address a variety of problems.

Research: Develop new ideas related to the above. Promisingideas make it into the bank’s codebase and are oftenintegrated by modelers or strats into production systems.





Hedge Funds/Prop Shops

Profile: Small companies: 10 - 1K employees, Small-MediumAUM: 50MM - 5B (typical) 10B - 100B (large).

Hedge Funds and Proprietary Trading shops differ from banksin that most of their revenue comes from trading with clientor house funds.

Less regulated than banks, are generally more risky from acareer perspective but typically have better upside potentialthan banks.

Quant Hedge Funds: Two Sigma, D.E. Shaw, KCG, Teza.

“Traditional” Hedge Funds: Duquesne Capital, Paulson &Co., Tiger Mgmt. Corp

Mixed Hedge Funds: Citadel, AQR, Bridgewater





Hedge Fund Quant Jobs

Mid-Frequency Algo Trading: Use math, statistics,optimization, and machine learning tools to work with tradersto find the best trading strategies subject to certainconstraints

Market Making: Post bid/ask prices for products that theytrade in and try to capture the spread. Have to manage thesize of inventory and consider trading market impact andexecution issues.

HFT: Execute trades at very high speeds (on the ordermicroseconds) and need quants who can implement relativelysimple algorithms in terms of model complexity in a highlyoptimized manner in say C/C++.





Fund Advisors/Asset Managers

Profile: Advise endowments, pension funds, corporations, andhigh net worth individuals on where to invest their money.Create their own mutual funds and ETFs.

Tend to have 100B - 3T in AUM but charge very low fees.

Asset Managers: Vanguard, PIMCO, Fidelity

Fund Advisors: Dimensional, Nuveen





Asset Manager Quant Jobs

Asset Allocation and Portfolio Construction: Variations of theMarkowitz portfolio, how to “tax optimize” portfolios.

Trade Execution: Understand potential market impact issueswhen executing large trades on behalf of clients (e.g. Wadell& Reed and the 2010 Flash Crash).

Risk/Hedging Quants: How does one minimize downside (i.e.“tail”) risk for different types of portfolios? How does thehedging effect the expected return of the portfolio?

Research Quants: Look into issues that are most relevant toclients sometimes on a case by case but more often on anaggregate basis and typically write results in the form ofnewsletters that are distributed to clients.




Risk MetricsThe GreeksRegulatory Requirements

Portfolio VaR Estimation

For a given portfolio, howmuch should one expect tolose on the worst day overa 20 or 100 trading dayperiod?

Addressing this probleminvolves estimating thequantiles of the returndistribution of theportfolio.

Use empirical, parametric,non-parametric methods.

What fundamental reasonscause the large losses?

−0.1 −0.05 0 0.05 0.10

20

40

60

80

100

120

140

160

180

200Daily Relative Returns of S&P 500 Survivors Over Last 7 Years, and 95% and 99% VaR

Daily Relative Return

Bin

Cou

nt





Portfolio Sensitivities to Market Variables

For a given portfolio, whatis the P&L impact whenunderlying market variableschange?

Common variables thattraders consider areinterest (LIBOR) rates, r ,credit (CDS spreads) c ,and portfolio volatility σ.

There are many ways toestimate these sensitivitiesto build a risk profile for aportfolio.

Linearize the portfolio’sprice:

P ≈∂P

∂rdr+

∂P

∂cdc+

∂P

∂σdσ+· · ·

Estimate the partials eitherusing a model or byempirical means.

Perform a scenarioanalysis, e.g. if r → r +∆r

and σ → σ − 2∆σ, thenwe have P → P +∆P ,what is ∆P?





Complying with Dodd-Frank

The Federal Reserve designed a new set of stress tests as partof the Dodd Frank Act to test whether or not banks willremain well capitalized during extreme economic environmentssuch as during the financial crisis.

To this point, only about half of the new Dodd-Frankmandates have been implemented at large bank holdingcompanies.

Failing stress tests or missing deadlines to implementregulatory requirements can result in increased regulatoryscrutiny or large fines for banks.

Risk Quants oversee the implementation of these requirementsand interact with regulators to explain their models andensure them that they satisfy external mandates.




Finding an Optimal StrategyUsing NLP to Construct Market SignalsMaking Markets

Best Mean Reverting Pair

Given a set of 1000 stocksprice time series, which twohave a spread that meanreverts more strongly thanany other pair?

What is the time scale ofthe mean reversion?

Does any stock’s pricemovements lead or laganother? What time scaledoes this lead/lagrelationship hold on? Daily,minute, second, shorter?

0 10 20 30 40 50 60 70 80 90 10070

80

90

100

110

120

130

Trading Days

Pric

e ($

)

Set of Price Time Series to Examine

0 10 20 30 40 50 60 70 80 90 10092

94

96

98

100

102

104

106

108

Trading Days

Pric

e ($

)

Two Series that Exhibit the Best Spread Mean Reversion Behavior

0 10 20 30 40 50 60 70 80 90 100−10

−5

0

5

10

15

Trading Days

Spr

ead

($)

Best Mean Reverting Spread





Using NLP to Predict Price Jumps

How does unexpected newsand quarterly reportinformation correspond tolarge moves in stockprices?

Can you develop softwarethat “reads the news”faster than a human?

What does it take to builda robust system to quicklyexecute trades on thesesignals?





Managing Inventory

Market makers provide liquidity by simultaneously posting bidand ask prices for a set of securities in which they specializeto try to capture the spread.

What levels do you set your bid/ask values at? They higherthe bid, the more likely you are to purchase a security andincrease your inventory.

The lower the ask, the more likely you are to sell which canimpact the market.

How can you dynamically update your bid/ask values toreflect your market views, minimize market impact, and run aprofitable business?




Skills to DevelopInterview Advice

Math/Stats Skills

Candidates with strong mathematics and statistics skills arehighly sought after by employers.

Math: Basic Probability Theory, Stochastic Processes, LinearAlgebra, PDE, Optimization.

Stats: Regression (linear, non-linear, non-parametric), DensityEstimation, Monte Carlo Simulation.

Machine Learning (Stats on Steroids): Classifers andRegression: SVMs, Neural Nets, k-means, etc.

Book References: Wasserman: All of Statistics, All ofNonparametric Statistics.

Software References: Python: scikit-learn, NumPy, SciPy.C++:Boost, Num. Recipes. MATLAB: Optimization, StatsToolbox.





Programming Skills

Quants will typically focus on either developing prototypes orwriting production quality code.

Prototyping Languages: Python (and supporting modules),MATLAB, R, Mathematica, and to a lesser degree Stata, SaS.

Production: C/C++, Java, C#, legacy Fortran systems.

Other: Linux/Unix, Excel/VBA, Git, SVN





Finance/Practical Skills

Develop an understanding of how math/stats methods can beapplied in practice.

Practice explaining things you learn in class to yournon-technical friends trying to get the high-level/big picturepoint across.

Book References: Bouchaud and Potters, Theory of FinancialRisk and Derivative Pricing. Hull, Options Futures and OtherDerivatives. McNeil, Frey, Embrechts. Quantitative RiskManagement.





Searching for a Job

Finding a Quantitative Finance Job is 50% Luck, 30% Skill,and 20% Timing!

The probability of getting any one job, PJ, is given by:

PJ = min [1, 0.5 U(0, 1) + 0.3 lnN (0, σSkill) + 0.2 N (0, σTime)]

Apply to as many jobs as possible that peek your interestsduring your job search.

Try to contact recruiters directly and ask them to forwardyour resume to the hiring manager. Applications throughwebsites of large corporations often are not routed to thecorrect people.

Try to do at least one internship during school.

Do not get discouraged, the process can take months andpersistence often pays off!





Before/After an Interview

Research the company you are speaking with prior to yourinterview and prepare questions relevant to the position youare seeking.

Make sure you focus on the needs of the company during theinterview and demonstrate how you can use past experienceyou have developed to solve their problems.

Do not BS the interviewer and do not overly exaggerate orembellish your past accomplishments.

Be ready to explain everything on your resume.

Write a thank you note to every person that interviews you.Hit on things that were discussed during the interview andreiterate your interest in the position. If they do not give youtheir contact info, send the notes directly to the recruiter.





Last Comments/The End

Use LinkedIn

Feel free to ask questions:

Contact Info: steve98654-at-gmail-dot-com

Thanks for Listening!


quantitative risk and trading examplesuserhome.brooklyn.cuny.edu/ohadjiliadis/talksteven.pdf ·...

Documents